AZ-900
Deep Dive
Public, private, hybrid, and multi-cloud. When to choose each one, which Azure services apply to each model, and the most common exam traps.
Contents
The AZ-900 tests 4 models. Many students only study 3 (public, private, hybrid) and lose points on multi-cloud questions.
Public Cloud
Infrastructure operated by a third-party cloud provider (Microsoft, AWS, Google) and delivered over the internet. Resources shared among multiple customers (multi-tenant) though logically isolated.
Examples
Best for
Most companies. Default for new projects.
Private Cloud
Cloud infrastructure used exclusively by one organization. Can be physically on the company's premises (on-premises) or in a third-party datacenter, but resources are NOT shared with others.
Examples
Best for
Ultra-strict compliance, predictable performance, total control.
Hybrid Cloud
A combination of public and private cloud (or on-premises) connected together, allowing data and applications to be shared between them. The application can "burst" to the public cloud when the private one becomes saturated.
Examples
Best for
Gradual migrations, compliance with burst capacity, legacy + new apps.
Multi-cloud
Deliberate use of services from two or more public cloud providers. Not the same as hybrid — it's multiple public clouds (e.g., Azure + AWS). Can also include private cloud.
Examples
Best for
Avoid vendor lock-in, take advantage of the best service from each provider.
What does "multi-tenant" actually mean?
Your VM in Azure runs on the same physical server as VMs from other companies (maybe competitors). The Hyper-V hypervisor isolates them completely at the hardware level. You can't access their data, nor can they access yours. But they share: CPU threads (limited by vCPU quota), physical memory (limited by allocated RAM), and the same physical NVMe disk (though with separate volumes).
Real advantages of multi-tenant
Risks and limitations of multi-tenant
Solution to noisy neighbor: Azure Dedicated Hosts
If you need exclusive physical hardware in public cloud → Azure Dedicated Hosts. You pay for the entire host, only your VMs run on it. Price: ~$3-8/hour for a complete physical host. Meets compliance requirements demanding non-shared hardware.
"Private cloud" doesn't just mean servers in your office. There are multiple possible architectures.
Traditional on-premises
Your own datacenter, your own servers, your own hypervisor (VMware, Hyper-V, OpenStack). Maximum control, maximum responsibility.
Advantages
Limitations
Relationship with Azure
You can extend Azure to your on-premises with Azure Arc
Azure Stack HCI
Microsoft-certified hardware that runs on your premises, managed like Azure from the Azure portal. You get Azure APIs and the Azure model, but the hardware is wherever you say.
Advantages
Limitations
Relationship with Azure
It's a Microsoft product — literally Azure in your datacenter
Colocation (colo)
Your hardware in a third party's datacenter (Equinix, NTT, Tier 4 providers). You own the hardware; the datacenter provides space, power, cooling, and connectivity.
Advantages
Limitations
Relationship with Azure
You can connect colo to Azure via ExpressRoute for private, high-speed connectivity
Hybrid cloud is the most common model in enterprises that have run on-premises infrastructure for years. Microsoft offers specific tools for this model.
Azure Arc — Extends Azure anywhere
Azure Arc lets you manage servers, Kubernetes clusters, SQL databases, and Azure services in any cloud or on-premises, from the Azure portal, with the same tools.
On-prem Linux/Windows servers
Appear in the Azure Portal as resources. Azure Monitor, Defender, Updates apply.
Kubernetes clusters on AWS/GCP
Managed from Azure. GitOps, Azure Policy applied.
On-premises SQL Server
Azure Arc-enabled SQL → telemetry, assessments, Defender for SQL.
Azure services anywhere
App Service, Kubernetes, Data Services on your hardware, managed by Azure.
Real hybrid cloud use cases
Hybrid connectivity in Azure
Azure VPN Gateway
Encrypted VPN connection over the internet. Economical, up to ~1 Gbps. For hybrid setups without strict SLA needs.
Azure ExpressRoute
Private dedicated connection (not over the internet) via a telecom partner. 1–100 Gbps. Guaranteed SLA. For critical workloads.
According to Flexera 2024, 89% of large enterprises use multiple clouds. Multi-cloud is no longer optional for many organizations.
Why do companies use multi-cloud?
Avoid vendor lock-in
Not depending on a single provider to negotiate better prices and not being vulnerable if the provider raises prices.
Best-of-breed per service
AWS Kinesis for streaming, Azure OpenAI for AI, Google BigQuery for analytics. Using the best service from each provider.
Corporate acquisitions
Company A uses Azure. It acquires Company B, which uses AWS. Multi-cloud by default.
Regulatory requirements
Some countries require using local or specific cloud providers for certain data.
Cross-provider redundancy
If Azure has a regional outage, failover to AWS. The probability of both failing simultaneously is extremely low.
Team with multiple expertise
A company has devs with AWS expertise and others with Azure. Each team uses what they know best.
The complexity of multi-cloud
Multi-cloud isn't free of complexity. Each cloud has its own: IAM model, CLI/APIs, networking, monitoring, compliance. You need teams that master multiple platforms. Egress costs between clouds can be significant. Governance becomes complex. Azure Arc can manage resources from other clouds from within Azure — one way to reduce this complexity.
| Dimension | Public | Private | Hybrid | Multi-cloud |
|---|---|---|---|---|
| Initial cost | 0 | Very high | Medium | 0 (per cloud) |
| Hardware control | None | Total | Mixed | None |
| Scalability | Unlimited | Limited to HW | Mixed | Unlimited (multi) |
| Operational management | Low | Very high | High | Very high |
| Long-term cost | Variable | Fixed (high) | Variable+fixed | Variable (multiple) |
| Physical security | Provider's | Yours | Mixed | Each provider's |
| Setup time | Minutes | 3-12 months | 2-6 months | Months (governance) |
| Vendor lock-in | High (1 vendor) | Your own | Partial | Low |
| Complexity | Low | High (yours) | High | Very high |
| Data residency compliance | Depends on region | Total control | Flexible | Multi-jurisdiction |
| Best for startups | ✓✓✓ | ✗ | ✗ | ✗ |
| Best for legacy enterprise | ✓ | ✓✓ | ✓✓✓ | ✓✓ |
Do you have regulatory requirements that mandate hardware PHYSICALLY on your premises?
Do you have legacy infrastructure you can't move to cloud in the short term?
Does your company already have significant investment in another cloud provider (AWS, GCP)?
Do you need specific services only offered by another provider (e.g., BigQuery, AWS Redshift)?
None of the above?
Did you get this topic?
Test what you just learned
A healthcare company needs to comply with strict regulations on where its data resides and cannot use shared infrastructure. Which cloud model should it choose?
Put what you learned to the test with original AZ-900 exam questions.
Start AZ-900 mock examUnlimited mock exams with the Pro plan
Log in to track your progress.